Wind farm station level weather forecasting method and system based on cloud edge end cooperation

Abstract

The embodiment of the specification provides a wind farm station-level weather prediction method and system based on cloud edge-end cooperation, wherein the method comprises: a cloud server performs data assimilation processing based on multi-source weather observation data, constructs an initial field, generates general low-resolution weather prediction data covering a target area based on the initial field, and sends the general low-resolution weather prediction data to an edge-end regional node; the edge-end regional node takes the low-resolution weather prediction data as initial conditions and boundary conditions of a mesoscale weather model in a weather research and forecast (WRF) model system, runs the WRF model to generate regional high-resolution weather prediction data, and sends the regional high-resolution weather prediction data to an end side deployed locally in a wind farm station; and the end side outputs a station-level refined weather prediction result based on the high-resolution weather prediction data through a trained calibration model.

Description

Technical Field

[0001] This document relates to the field of wind power meteorological forecasting technology, and in particular to a wind farm-level meteorological forecasting method and system based on cloud-edge-device collaboration. Background Technology

[0002] The accuracy of wind power forecasting hinges on the precision of weather forecasts. Currently, weather forecasting for wind farms faces the following main challenges: (1) Traditional meteorological numerical forecasts (such as GFS, EC, etc.) have low spatial resolution (usually 10-25km), which cannot meet the needs of fine forecasting at the field station level (hundred-meter level). If high-resolution numerical forecasting models (such as WRF) are run directly in the cloud, the computational resources consumed are huge and the cost is unbearable, especially in scenarios that need to serve a large number of dispersed wind farms.

[0003] (2) The terrain, landforms and microclimate conditions of each wind farm are significantly different, and the general model is difficult to adapt to the local characteristics of all sites, resulting in insufficient prediction accuracy.

[0004] (3) Existing solutions typically employ either pure cloud-based centralized computing or pure local deployment. The former is costly and slow to respond; the latter requires deploying a complete forecasting system at each site, which is complex to operate and maintain and makes it difficult to share global meteorological information.

[0005] In other words, achieving high-precision weather forecasting at the wind farm level while controlling computational costs, and adapting to the local meteorological characteristics of different sites, is a pressing technical problem that needs to be solved. Therefore, there is an urgent need for a wind farm-level weather forecasting solution that balances accuracy, cost, and scalability. Summary of the Invention

[0006] The purpose of this invention is to provide a wind farm-level meteorological forecasting method and system based on cloud-edge-device collaboration, aiming to solve the above-mentioned problems in the prior art.

[0007] This invention provides a wind farm-level meteorological forecasting method based on cloud-edge-device collaboration, comprising: The cloud server performs data assimilation processing based on multi-source meteorological observation data, constructs an initial field, generates general low-resolution meteorological forecast data covering the target area based on the initial field, and sends it to the edge area nodes. The edge regional nodes use the low-resolution meteorological forecast data as the initial and boundary conditions of the scale meteorological model in the Weather Research and Forecasting (WRF) model system, run the WRF model to generate regional high-resolution meteorological forecast data, and send it to the edge side deployed locally at the wind farm station. Based on the high-resolution meteorological forecast data, the terminal outputs refined meteorological forecast results at the station level through the trained calibration model.

[0008] This invention provides a wind farm-level meteorological forecasting system based on cloud-edge-device collaboration, comprising: The cloud-based prediction module is located on a cloud server and is used to perform data assimilation processing based on multi-source meteorological observation data, construct an initial field, generate general low-resolution meteorological prediction data covering the target area based on the initial field, and send it to the edge area nodes. The edge downscaling module, located at the edge regional node, is used to take the low-resolution meteorological forecast data as the initial and boundary conditions of the scale meteorological model in the WRF model system for weather research and forecasting, run the WRF model to generate regional high-resolution meteorological forecast data and send it to the edge side deployed locally at the wind farm station. An end-side calibration module, located at the end, is used to output refined weather forecast results at the station level based on the high-resolution meteorological forecast data and the trained calibration model.

[0009] This invention also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the above-described wind farm-level meteorological forecasting method based on cloud-edge-device collaboration.

[0010] This invention also provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor, implements the steps of the above-described wind farm-level meteorological forecasting method based on cloud-edge-device collaboration.

[0011] By employing this invention, a three-tiered architecture is used to rationally distribute the computational load of high-precision prediction across the cloud, edge, and end-user devices, significantly reducing the computational cost of pure cloud-based high-resolution prediction. The edge-user device uses WRF dynamic downscaling to ensure physical consistency and achieves higher prediction accuracy in complex terrains compared to statistical downscaling methods. The end-user device utilizes historical data from local sensors at the site to train calibration models, adapting to the unique microclimate characteristics of each site and solving the problem of insufficient accuracy in general models. The end-user model has an automatic monitoring and retraining mechanism, enabling it to adapt to changes in site meteorological characteristics and ensure long-term stability of prediction accuracy. Data flows unidirectionally, with end-user data not transmitted back to the cloud, ensuring the security and privacy of site data. The architecture has excellent scalability; adding a new site only requires deploying a calibration module on the end-user device and connecting to edge data, without needing to adjust the cloud and edge architecture. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a wind farm-level meteorological forecasting method based on cloud-edge-device collaboration according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the system architecture of the wind farm-level meteorological forecasting method based on cloud-edge-device collaboration according to an embodiment of the present invention; Figure 3 This is a schematic diagram of data flow in the wind farm-level meteorological forecasting method based on cloud-edge-device collaboration according to an embodiment of the present invention; Figure 4 This is a schematic diagram of edge model monitoring and automatic retraining in an embodiment of the present invention; Figure 5 This is a schematic diagram of a wind farm-level meteorological forecasting system based on cloud-edge-device collaboration, according to an embodiment of the present invention. Figure 6 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0015] Method Implementation Examples According to embodiments of the present invention, a wind farm-level meteorological forecasting method based on cloud-edge-device collaboration is provided. Figure 1 This is a flowchart of a wind farm-level meteorological forecasting method based on cloud-edge-device collaboration, as described in an embodiment of the present invention. Figure 1 As shown, the wind farm-level meteorological forecasting method based on cloud-edge-device collaboration according to an embodiment of the present invention specifically includes: Step S101: The cloud server performs data assimilation processing based on multi-source meteorological observation data to construct an initial field. Based on the initial field, it generates general low-resolution meteorological forecast data covering the target area and sends it to the edge region nodes; specifically including: The cloud server performs data assimilation processing based on multi-source meteorological observation data, constructs an initial field, and runs a low-resolution meteorological numerical forecasting model. The low-resolution meteorological numerical forecasting model has a low resolution of greater than or equal to 25 km. The multi-source meteorological observation data includes at least one of the following: ground observation station data, radiosonde data, satellite remote sensing data, and radar data. The low-resolution meteorological numerical forecast model generates general low-resolution meteorological forecast data covering the target area, wherein the low-resolution meteorological forecast data includes the spatiotemporal distribution forecasts of meteorological elements such as wind speed, wind direction, temperature, air pressure, and humidity.

[0016] Step S102: The edge regional node uses the low-resolution meteorological forecast data as the initial and boundary conditions of the scale meteorological model in the Weather Research and Forecasting (WRF) model system, runs the WRF model to generate regional high-resolution meteorological forecast data, and sends it to the edge side deployed locally at the wind farm station; specifically including: Edge regional nodes receive low-resolution meteorological forecast data output from the cloud and use this low-resolution meteorological forecast data as the initial and boundary conditions of the scale meteorological model in the WRF model system. They then run the WRF model to perform dynamic downscaling calculations, improving the forecast resolution from low to high resolution and generating high-resolution regional meteorological forecast data. The edge regional nodes are deployed according to geographical regions, and each edge regional node is responsible for covering multiple wind farms within a certain range. The high resolution is 1-3km.

[0017] Step S103: Based on the high-resolution meteorological forecast data, the terminal outputs station-level refined meteorological forecast results through the trained calibration model.

[0018] In step 103, the edge device needs to train a station-level prediction calibration model based on the high-resolution meteorological forecast data and the historical sensor observation data of the current station. This prediction calibration model is used to learn the systematic deviation between the edge prediction values ​​and the actual station observation values, and to perform localized calibration of the edge prediction results. Furthermore, the edge device continuously monitors the prediction accuracy of the prediction calibration model. When the deviation between the predicted value and the actual observation value exceeds a preset threshold, the retraining process of the prediction calibration model is automatically triggered. In the retraining process, the parameters of the prediction calibration model are updated using the latest accumulated historical sensor data and the corresponding edge prediction data. The preset threshold is the root mean square error threshold between the predicted value and the actual observation value.

[0019] The technical solutions of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] This invention proposes a wind farm-level meteorological forecasting method based on cloud-edge-device three-level collaboration, the system architecture of which is as follows: Figure 2 As shown in the diagram, the data flow is illustrated below. Figure 3 As shown, the specific steps include: Step 1: Cloud-based general meteorological numerical forecasting The cloud server performs data assimilation processing based on multi-source meteorological observation data to construct an initial field, runs a low-resolution (e.g., 25km or lower) meteorological numerical forecast model, and generates general low-resolution meteorological forecast data covering the target area. The multi-source meteorological observation data includes, but is not limited to, ground-based observation station data, radiosonde data, satellite remote sensing data, and radar data. The low-resolution meteorological forecast data includes spatiotemporal distribution predictions of meteorological elements such as wind speed, wind direction, temperature, air pressure, and humidity.

[0021] Step 2: Downscaling of edge regions Edge nodes receive low-resolution weather forecast data output from the cloud, using this as the initial and boundary conditions for the WRF (Weather Research and Forecasting) mesoscale weather model. The WRF model is then run to perform dynamic downscaling calculations, increasing the forecast resolution from low to high (e.g., 1-3 km), generating high-resolution regional weather forecast data. These edge nodes can be deployed geographically, with each node responsible for covering multiple wind farms within a certain area.

[0022] Step 3: End-side station model debugging and refined prediction The edge-side system is deployed locally at the wind farm site to receive high-resolution meteorological forecast data output from the edge-side system. Combined with historical sensor observation data from the site, a site-level forecast calibration model is trained. This calibration model is used to learn the systematic deviations between the edge-side forecast values ​​and the actual observation values ​​at the site, performing localized calibration of the edge-side forecast results, and outputting refined meteorological forecast results at the site level.

[0023] Step 4: On-device model monitoring and automatic retraining, such as... Figure 4 As shown: The accuracy of the calibration model is continuously monitored at the edge. When the deviation between the predicted value and the actual observed value exceeds a preset threshold (e.g., the root mean square error exceeds a set proportion), the model retraining process is automatically triggered. Retraining uses the latest accumulated historical sensor data and corresponding edge prediction data to update the calibration model parameters to adapt to the seasonal or long-term changes in the meteorological characteristics of the stations.

[0024] In this method, the data flow is unidirectional: cloud → edge → terminal. The terminal does not return data to the cloud, ensuring localized processing and data security of the site data.

[0025] In summary, the present invention has the following beneficial effects: (1) By using a three-level architecture for layered processing, the computational load of high-precision prediction is reasonably distributed to the cloud, edge and end, which significantly reduces the computational cost of pure cloud high-resolution prediction. (2) The WRF dynamic downscaling method is used at the edge to ensure physical consistency and has higher prediction accuracy in complex terrain compared with the statistical downscaling method. (3) The end side uses historical data from local sensors at the station to train and calibrate the model, which can adapt to the unique microclimate characteristics of each station and solve the problem of insufficient accuracy of the general model. (4) The end-side model has an automatic monitoring and retraining mechanism, which can adapt to changes in the meteorological characteristics of the station and ensure the long-term stability of the prediction accuracy. (5) Data flows in one direction only, and the data on the terminal side is not transmitted back to the cloud, which ensures the data security and privacy of the site; (6) The architecture has good scalability. When adding a new site, you only need to deploy the calibration module on the end side and connect the edge data. There is no need to adjust the cloud and edge architecture.

[0026] System Implementation Examples According to embodiments of the present invention, a wind farm-level meteorological forecasting system based on cloud-edge-device collaboration is provided. Figure 5 This is a schematic diagram of a wind farm-level meteorological forecasting system based on cloud-edge-device collaboration, as described in an embodiment of the present invention. Figure 5 As shown, the wind farm-level meteorological forecasting system based on cloud-edge-device collaboration according to an embodiment of the present invention specifically includes: The cloud prediction module 50 is set on the cloud server and is used to perform data assimilation processing based on multi-source meteorological observation data, construct an initial field, generate general low-resolution meteorological prediction data covering the target area based on the initial field, and send it to the edge area node. The edge downscaling module 52 is set at the edge region node and is used to take the low-resolution meteorological forecast data as the initial and boundary conditions of the scale meteorological model in the WRF model system for weather research and forecasting, run the WRF model to generate regional high-resolution meteorological forecast data and send it to the edge side deployed locally at the wind farm station. The end-side calibration module 54 is located on the end side and is used to output station-level refined meteorological forecast results based on the high-resolution meteorological forecast data and the trained calibration model.

[0027] The following are examples, such as Figure 2 As shown: (1) Cloud-based forecasting module: Deployed on the central server, including a data assimilation submodule and a low-resolution numerical forecasting submodule, used to generate general low-resolution meteorological forecast data; (2) Edge downscaling module: Deployed at the edge nodes of the region, including the WRF dynamic downscaling calculation submodule, which receives low-resolution forecast data from the cloud and outputs high-resolution meteorological forecast data for the region; (3) End-side calibration module: Deployed locally at the wind farm, including a site calibration model training submodule, a prediction calibration submodule and a model monitoring and retraining submodule, receiving high-resolution prediction data from the edge and performing localized calibration by combining local sensor historical data; (4) Data transmission channel: connects the cloud, edge and end, and supports one-way data transmission from cloud to edge and from edge to end.

[0028] The embodiments of the present invention are system embodiments corresponding to the above method embodiments. The specific operation of each module can be understood by referring to the description of the method embodiments, and will not be repeated here.

[0029] Device Example 1 This invention provides an electronic device, such as... Figure 6 As shown, it includes: a memory 60, a processor 62, and a computer program stored in the memory 60 and executable on the processor 62, wherein the computer program, when executed by the processor 62, performs the steps as described in the method embodiment.

[0030] Device Example 3 This invention provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor 62, performs the steps described in the method embodiment.

[0031] The computer-readable storage media described in this embodiment include, but are not limited to, ROM, RAM, disk, or optical disk.

[0032] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A wind farm-level meteorological forecasting method based on cloud-edge-device collaboration, characterized in that, include: The cloud server performs data assimilation processing based on multi-source meteorological observation data, constructs an initial field, generates general low-resolution meteorological forecast data covering the target area based on the initial field, and sends it to the edge area nodes. The edge regional nodes use the low-resolution meteorological forecast data as the initial and boundary conditions of the scale meteorological model in the Weather Research and Forecasting (WRF) model system, run the WRF model to generate regional high-resolution meteorological forecast data, and send it to the edge side deployed locally at the wind farm station. Based on the high-resolution meteorological forecast data, the terminal outputs refined meteorological forecast results at the station level through the trained calibration model.

2. The method according to claim 1, characterized in that, The method further includes: Based on the high-resolution meteorological forecast data and combined with the historical sensor observation data of the current station, the edge side trains a station-level prediction calibration model. The prediction calibration model is used to learn the systematic deviation between the edge prediction value and the actual observation value of the station, and to perform localized calibration of the edge prediction results.

3. The method according to claim 1, characterized in that, The method further includes: The device continuously monitors the prediction accuracy of the prediction calibration model. When the deviation between the predicted value and the actual observed value exceeds a preset threshold, the retraining process of the prediction calibration model is automatically triggered.

4. The method according to claim 2, characterized in that, The method further includes: The parameters of the prediction calibration model are updated using the latest accumulated historical sensor data and corresponding edge prediction data during the retraining process.

5. The method according to claim 1, characterized in that, The cloud server performs data assimilation processing based on multi-source meteorological observation data to construct an initial field. Based on the initial field, it generates general low-resolution meteorological forecast data covering the target area and sends it to the edge area nodes. Specifically, this includes: The cloud server performs data assimilation processing based on multi-source meteorological observation data, constructs an initial field, and runs a low-resolution meteorological numerical forecasting model. The low-resolution meteorological numerical forecasting model has a low resolution of greater than or equal to 25 km. The multi-source meteorological observation data includes at least one of the following: ground observation station data, radiosonde data, satellite remote sensing data, and radar data. The low-resolution meteorological numerical forecast model generates general low-resolution meteorological forecast data covering the target area, wherein the low-resolution meteorological forecast data includes the spatiotemporal distribution forecasts of meteorological elements such as wind speed, wind direction, temperature, air pressure, and humidity.

6. The method according to claim 1, characterized in that, The edge regional nodes use the low-resolution meteorological forecast data as the initial and boundary conditions for the large-scale meteorological model in the Weather Research and Forecasting (WRF) model system, run the WRF model to generate regional high-resolution meteorological forecast data, and send it to the edge nodes deployed locally at wind farms. Specifically, this includes: Edge regional nodes receive low-resolution meteorological forecast data output from the cloud and use this low-resolution meteorological forecast data as the initial and boundary conditions of the scale meteorological model in the WRF model system. They then run the WRF model to perform dynamic downscaling calculations, improving the forecast resolution from low to high, and generating high-resolution regional meteorological forecast data. The edge regional nodes are deployed according to geographical regions, and each edge regional node is responsible for covering multiple wind farms within a certain range. The high resolution is 1-3km.

7. The method according to claim 3, characterized in that, The preset threshold is the root mean square error threshold between the predicted value and the actual observed value.

8. A wind farm-level meteorological forecasting system based on cloud-edge-device collaboration, characterized in that, include: The cloud-based prediction module is located on a cloud server and is used to perform data assimilation processing based on multi-source meteorological observation data, construct an initial field, generate general low-resolution meteorological prediction data covering the target area based on the initial field, and send it to the edge area nodes. The edge downscaling module, located at the edge regional node, is used to take the low-resolution meteorological forecast data as the initial and boundary conditions of the scale meteorological model in the WRF model system for weather research and forecasting, run the WRF model to generate regional high-resolution meteorological forecast data and send it to the edge side deployed locally at the wind farm station. An end-side calibration module, located at the end, is used to output refined weather forecast results at the station level based on the high-resolution meteorological forecast data and the trained calibration model.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the wind farm-level meteorological forecasting method based on cloud-edge-device collaboration as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an implementation program for information transmission, which, when executed by a processor, implements the steps of the wind farm-level meteorological forecasting method based on cloud-edge-device collaboration as described in any one of claims 1 to 7.

Images